Friday, 08 March 2019, 14:00 at MR11, Centre for Mathematical Sciences, Wilberforce Road, Cambridge – Classical nonlinear optimisation algorithms require the availability of gradient evaluations for constructing local approximations to the objective and testing for convergence. In settings where the objective is expensive to evaluate or noisy, evaluating the gradient may be too expensive or inaccurate, so cannot be used; we must turn to optimisation …